Hyperactivity of Rac1-GTPase pathway impairs neuritogenesis of cortical neurons by altering actin dynamics

The small-GTPase Rac1 is a key molecular regulator linking extracellular signals to actin cytoskeleton dynamics. Loss-of-function mutations in RAC1 and other genes of the Rac signaling pathway have been implicated in the pathogenesis of Intellectual Disability (ID). The Rac1 activity is negatively controlled by GAP proteins, however the effect of Rac1 hyperactivity on neuronal networking in vivo has been poorly studied. ArhGAP15 is a Rac-specific negative regulator, expressed in the main subtypes of pyramidal cortical neurons. In the absence of ArhGAP15, cortical pyramidal neurons show defective neuritogenesis, delayed axonal elongation, reduced dendritic branching, both in vitro and in vivo. These phenotypes are associated with altered actin dynamics at the growth cone due to increased activity of the PAK-LIMK pathway and hyperphosphorylation of ADF/cofilin. These results can be explained by shootin1 hypo-phosphorylation and uncoupling with the adhesion system. Functionally, ArhGAP15−/− mice exhibit decreased synaptic density, altered electroencephalographic rhythms and cognitive deficits. These data suggest that both hypo- and hyperactivation of the Rac pathway due to mutations in Rac1 regulators can result in conditions of ID, and that a tight regulation of Rac1 activity is required to attain the full complexity of the cortical networks.


Results
Expression of ArhGAP15 in the postnatal mouse brain. We previously showed that in the early postnatal brain, ArhGAP15 is expressed in the olfactory bulbs, cortex and hippocampus 18 . Xgal staining of newborn (P1) ArhGAP15 +/− brains showed abundant βgal-expressing cells in the somatosensory and somatomotor cortical areas (Fig. 1A,B). We characterized the expression in pyramidal neurons by double immunostaining of ArhGAP15 +/− brains at the age P1 with anti-βgal (detecting ArhGAP15-expressing cells) and either Ctip2, Cux1 or Tbr1 (markers, respectively, of pyramidal neurons in layer V, layers II-III and layer VI). The use of anti-βgal is justified by the fact that existing antibodies directed against ArhGAP15 also detect unspecific signal, and cannot be reliably used. βgal/ArhGAP15 was detected nearly exclusively in Ctip2+ neurons in layer V, while it was not detected in Cux1+ and Tbr1+ neurons in other layers ( Fig. 1C-G). Quantification of the double-stained neurons indicated that only a fraction (40% ± 3.84%) of the Ctip2+ neurons is βgal+ (Fig. 1O).
In young adults (P60) we carried out RNA::RNA in situ hybridization on coronal sections of WT brains, to detect the ArhGAP15 mRNA; expression was observed in the olfactory bulbs, cortex and hippocampus (Fig. 1H). Similarly, in situ hybridization with a probe detecting the lacZ mRNA on ArhGAP15 +/− brains showed an identical expression pattern. We then carried out Xgal staining of coronal sections of P60 ArhGAP15 +/− brains and again we observed expression of βgal in all regions of the cerebral cortex as well as the hippocampus (Fig. 1I). Finally, we carried out Xgal staining and the lacZ RNA::RNA in situ hybridization to compare ArhGAP15 +/− and ArhGAP15 −/− brains, and observed no difference between the two genotypes ( Supplementary Fig. S1A,B).
To examine the expression of ArhGAP15 in young adult pyramidal neurons in more details, we carried out double immunostaining on ArhGAP15 +/− brains at the age P45, to detect βgal and either Ctip2, Cux1 and Tbr1, markers of layer-specific and target-specific subclasses of cortical pyramidal neurons. In the adult ArhGAP15 +/− brain βgal/ArhGAP15 was detected in a fraction of all pyramidal neuron subtypes (Ctip2/βgal 43% ± 4.80%; Cux1/βgal24% ± 2.38%; Tbr1/βgal 28% ± 7.45%) (Fig. 1J-O). In summary ArhGAP15 is expressed postmitotically, in a time-dependent manner, in a fraction of most subtypes of cortical pyramidal neurons, the first ones being Ctip2+ neurons in layer V.
Cortical organization is normal in ArhGAP15 −/− mice. Xgal staining and lacZ RNA::RNA in situ hybridization showed no evident difference between normal and ArhGAP15 −/− brains in terms of distribution of the reporter-expressing cells ( Supplementary Fig. S1A,B), indicating that the ArhGAP15/lacZ expressing cells are not lost or grossly mislocalized in the absence of ArhGAP15. In addition, immunostaining of P1 brain sections for Ctip2 showed no difference in the neuron number and position ( Supplementary Fig. S1C,D). Furthermore, staining of coronal sections of the early post-natal or adult animals with TUNEL for the detection of apoptotic cells revealed no significant change in the number and position of TUNEL+ cells, thus we exclude a loss of specific cell populations.
To better evaluate the genesis and cyto-architecture of cortical pyramidal neurons in the absence of ArhGAP15, we pulse-labeled proliferating neuroblasts with a single injection of EdU at E14.5 and then determined the  number and distribution of EdU+, Satb2+ and Tbr1+ neurons detected at birth. Satb2 is a postmitotic determinant for upper-layer cortical neurons 25 , while Tbr1 is preferentially expressed in layer VI neurons 26 . We observed no difference in either the laminar organization of Satb2 or Tbr1 pyramidal neurons or in the number/position of EdU-labeled neurons between the two genotypes ( Fig. 2). We conclude that, at birth, in the absence of ArhGAP15 pyramidal neurons are normally generated and positioned.

Loss of ArhGAP15 affects neuronal morphology and neuritogenesis in vitro and in vivo.
The small-GTPase Rac1 is critical for the acquisition of neuronal morphology and for neuritogenesis during development 27 . To determine whether loss of ArhGAP15, hence hyperactive Rac1/Rac3, affects neuronal morphology, neurite elongation, branching and overall neurite complexity, we examined primary cultures of neurons dissociated from cortices of WT and ArhGAP15 −/− embryos, at the age E15.5, electroporated with a Green Fluorescent Protein (GFP) expressing vector and maintained 3 DIV ( Supplementary Fig. S2A). First, we categorized neurons on the basis of the number of main neurites, and classified them as unipolar (having one visible neurite), bipolar (two neurites) and multipolar (>two neurites) (Supplementary Fig. S2B-D). We found that the fraction of unipolar cells was significantly increased (WT 5% vs. KO 12%) in cultures of ArhGAP15 −/− cortices, while the fraction of bipolar cells was diminished (WT 21% vs. KO 14%) (Fig. 3A). The fraction of multipolar neurons was unchanged in the absence of ArhGAP15 −/− , however the number of their neurites is reduced in ArhGAP15 −/− neurons, compared to WT (WT 6 ± 0.18 vs. KO 5 ± 0.23) (Fig. 3B). These data indicate that ArhGAP15 is required by immature neurons to achieve a more elaborated morphology, and in the absence of ArhGAP15 the formation of the first extension might be delayed.
We then examined neurite length and complexity of pyramidal-looking neurons (which represent >90% of cultured neurons). ArhGAP15 −/− neurons displayed a reduced length of the main neurite, a reduced number of secondary neurites, and reduced intersection distribution and number of intersections (respectively, p = 2.3 10 −4 ; 9.4 10 −7 and 4.5 10 −11 ) (Fig. 3C-E). Moreover, Sholl analysis revealed that the ArhGAP15 −/− pyramidal neurons display nearly one-third of the number of neurites stemming from the soma and a reduced number of branches crossing more distal concentric circles centered at the cell body, compared to the WT (Fig. 3E). These data indicate that in the absence of ArhGAP15 neurons possess a simpler morphology and reduced efficiency of neuritogenesis and branching, in vitro.
To further investigate the axonal elongation and the dendrite branching, we evaluated the neuronal morphology and neurite length in vivo. We implanted crystals of DiI in the corpus callosum of sections of P2 brains from WT (N = 10 from 5 brains) and ArhGAP15 −/− (N = 10 from 5 brains). After 7 days of diffusion time, we examined the dendritic complexity of DiI-labelled callosal neurons (Fig. 3F,H) using the Neurolucida software (Fig. 3G,I). In the ArhGAP15 −/− brains we detected a reduced length of axons (WT 0.22 mm ± 0.03 vs. KO 0.11 mm ± 0.01) and dendrites (WT 0.22 mm ± 0.03 vs. KO 0.08 mm ± 0.02) (Fig. 3J). Moreover Sholl analysis indicates that ArhGAP15 −/− callosal neurons display a less complex neuronal arbor (Fig. 3K). These data indicate that ArhGAP15 is required by neurons to attain a mature morphology and a complex dendritic and axonal arborization.   (N = 7 from 3 brains) and ArhGAP15 −/− (N = 12 from 4 brains) animals, and examined the length and organization of DiI-labelled projections (callosal) to the contralateral side, after 7 days of diffusion time (Fig. 4A-D). While the general organization was unchanged, in the ArhGAP15 −/− brains we observed that the median length of DiI+ axons is reduced (−15%, p < 0.001) and the number of fibers reaching the maximum length is reduced (−20%, p < 0.001) (Fig. 4E). In the same sections, we also determined the average length (distance from the midline) of the 5 longest axons, on a minimum of 50 total axons, and found that their length is significantly reduced in the absence of ArhGAP15 (Fig. 4F). We repeated this analysis at P8 and examined the length of DiI-labelled projection in the contralateral side; also at this age we observed a significant delay.

Loss of
Next we examined the length and targeting of callosal axons at P45, when the axons have already reached their target. We traced callosal axons length by stereotactically injecting BDA into the right cortex of WT and ArhGAP15 −/− animals, and examining the contralateral callosal projection 7 days later, by avidin-biotinylated HRP staining. In the two genotypes, BDA+ contralateral callosal projection reached equal distances, as no significant difference was detected ( Fig. 4G-K). Collectively, these results indicate that in the absence of ArhGAP15 pyramidal callosal axons show delayed elongation in vivo.

Retrograde actin flow is increased in the absence of
ArhGAP15. The small-GTPase Rac1 acts by controlling actin cytoskeleton dynamics at the axonal growth cone, as well as at the leading edge of migrating neuroblasts 27 . Therefore we examined whether the observed inefficient neuritogenesis might be linked to altered actin dynamics at the growth cone of ArhGAP15 −/− neurons. We established primary cultures of cortical neurons, and electroporated the dissociated cells with RFP-Lifeact 28 , a vector expressing a protein conjugated with RFP and able to bind dynamic actin (a strategy widely used to visualize sites of active actin polymerization/depolymerization in neurons) ( Supplementary Fig. S3A,B). Nearly 100% cells stained positive for βIII-tubulin, an early pan-neuronal marker ( Supplementary Fig. S3C). Cultures were composed of approximately 92% pyramidal neurons, since staining of both WT and ArhGAP15 −/− cultures with anti-GAD67 detected 8% of GAD67+ neurons ( Supplementary Fig. S3D,E).
After 2 DIV, the RFP+ axonal growth cones were subjected to time-lapse video recording, and the raw data were statistically analysed with ImageJ and the dedicated plugins. Considering only the growth cone of the longest neurite (most likely the axon) of each recorded neuron, we observed that in the absence of ArhGAP15 the retrograde actin movements were significantly increased, while the anterograde flow was not significantly changed ( Fig. 5A,B and Supplementary Movies S1 and S2). A significant difference was detected also when we considered separately neurons with a pyramidal-like morphology (based on the cell body) and those (few ones) with a non-pyramidal morphology. Instead, when we examined the growth cones of secondary neurites, no significant alteration was observed. These data strongly indicate that growth cone actin dynamics is altered in ArhGAP15 −/− neurons, suggesting an inverse relationship between actin retrograde flow and axon elongation.

Loss of ArhGAP15 results in increased cofilin phosphorylation via the PAK-LIMK pathway.
We have previously shown that the genetic disruption of ArhGAP15 leads to complete absence of the ArhGAP15 protein and that, as consequence of the absence of ArhGAP15, Rac1 activity is 1.7-fold (E15.5) and a 2-fold (P2) increased 18,23 . Rho-GTPases are linked to the dynamics of actin cytoskeleton via the control of the actin-binding and actin-severing phosphoprotein ADF/cofilin. We therefore examined the levels of phospho-cofilin in protein extracts from WT or ArhGAP15 −/− neonatal cortices (P2), by Western blot analyses. In samples from ArhGAP15 −/− cortices, we observed a significant increase of the phospho-cofilin/total cofilin ratio, while the total cofilin level was unchanged (Fig. 5C). Considering that one of the best studied pathways involved in the phosphorylation (and inactivation) of ADF/cofilin is the Rac-PAK-LIMK 29,30 , we examined the phosphorylation state of LIMK1/2 and PAK1-3. In protein samples from ArhGAP15 −/− cortices both LIMK1/2 and PAK1-3 were significantly hyperphorphorylated (Fig. 5D). Finally we examined the activation state of slingshot-1L (SSH1L), an ADF/cofilin phosphatase which is activated by phosphorylation. In protein samples from ArhGAP15 −/− cortices we observed a significant reduction in SSH1L phosphorylation (Fig. 5D), indicating that this phosphatase is hypoactive. In conclusion, Rac1 hyperactivity consequent to the loss of ArhGAP15, results in ADF/cofilin inactivation via the increased activity of the PAK-LIMK pathway, and reduced activity of the SSH1L phosphatase.
In order to link the increased ADF/cofilin phosphorylation (hence inactivation) with the increased retrograde actin flow and with shorter and less complex neurites in ArhGAP15 −/− neurons, we reasoned that loss of ArhGAP15 may result in decreased phosphorylation of shootin1 and reduced coupling with L1 adhesion. Indeed, increased retrograde flow of the actin network may result from the combined effect of higher actin polymerization and increased myosin II motor function not accompanied by active association of actin-binding proteins with adhesion complexes 9 . Recently, shootin1 phosphorylation downstream of Rac1/Rac3 has been recognized as a key mechanism to couple enhanced actin flow with increased efficacy of adhesion on L1 substrates, needed for actual neurite elongation 31,32 . We examined the phosphorylation state of shootin1 in protein samples from WT and ArhGAP15 −/− cortices, by Western blot analyses. We observed a significant reduction in the phosphorylation state of shootin1 in ArhGAP15 −/− samples (Fig. 5D). With inactive shootin1, thus an altered engagement of the "clutch" system linking actin filament elongation with L1 dependent adhesion, explains both the increased retrograde actin flow and the inefficient neurite elongation of ArhGAP15 null neurons.

Loss of ArhGAP15 alters synaptic density and causes hyperexitability. Given that Rac GTPases
and their regulators have been extensively implicated in early steps of spinogenesis and synaptogenesis 33 , and considering the inefficient neuritogenesis and branching of ArhGAP15 −/− neurons, we set forth to examine the number of excitatory (VGLUT+) and inhibitory (VGAT+) synapses on the soma of the pyramidal cortical neurons. We determined the density of VGLUT+ and VGAT+ punctae on the perisomatic surface of pyramidal neurons in layer III of the somatosensory cortex of WT and ArhGAP15 −/− mice at the age P45. We observed a significant reduction (−22.5% in KO vs. WT) of the density of VGLUT+ punctae per equal area (81.9 μm 2 ) and a significant reduction of the density of VGAT+ punctae on the pyramidal soma (WT 13/soma ± 0.38 vs. KO 10/ soma ± 0.2) (WT 0.41/μm ± 0.023 vs.0.35/μm ± 0.012) (Fig. 6A-F). This result suggests that reduced efficiency of neuritogenesis is associated with an overall reduced efficiency of synaptogenesis onto cortical pyramidal neurons.
To evaluate the functional significance of this, we examined the passive and active electrophysiological properties of cortical pyramidal neurons, comparing brain slices from ArhGAP15 −/− animals with those from control WT ones. We performed current-clamp recordings from pyramidal neurons in the somatosensory cortex, identified on the basis of their morphological features, of three ArhGAP15 −/− and three WT males, at the age P100. The passive membrane properties, including capacitance, input resistance and resting potential were unchanged in the ArhGAP15 −/− neurons compared with the WT controls ( Fig. 6G-J). On the contrary, in the absence of ArhGAP15 the action potential threshold was diminished (WT −43.5 ± 1.0 mV, KO −47.7 ± 0.8 mV, p < 0.015) and the action potential frequency at 200 pA of injected current was increased (WT 21.8 ± 0.7 Hz; KO 26.9 ± 2.8 Hz, p < 0.045) in neurons from ArhGAP15 −/− animals (Fig. 6K-N). These results indicate that cortical pyramidal neurons lacking ArhGAP15 are more excitable compared to WT neurons. Western blot analyses of ADF/cofilin and phospho-ADF/cofilin in total extracts of WT and ArhGAP15 −/− cortices, at P2. The signal relative to GAPDH is used as loading control and subsequent normalization. On the right, quantification of the ratio of phospho-ADF/cofilin over the amount of total ADF/cofilin and phospho-ADF/cofilin, corrected for GAPDH expression. The WT is placed = 1. (D) Western blot analyses of phospho-PAK1-3, phospho-LIMK1/2, phospho-SSHL1 and phospho-shootin1 in total extracts of WT and ArhGAP15 −/− cortices, at P2. The signal relative to GAPDH is used as loading control and subsequent normalization. On the right, quantification of the ratio of phospho-PAK1-3, phospho-LIMK1/2, phospho-SSHL1 and phospho-shootin1 corrected for GAPDH expression. The WT is considered = 1. Full-length blots are presented in Supplementary Figure S4. Data are presented as mean ± SEM. *P ≤ 0.05 (Student's t-test).  Loss of ArhGAP15 alters the EEG oscillation spectrum. To evaluate the functional significance of the previous findings, we examined the global electrical activities in the intact brain. In particular, imbalance between excitation and inhibition correlates with changes in power of γ-range oscillations 34,35 and reduced neuronal complexity seen in mouse models of Huntington Disease is associated with low-shifts of the mean dominant frequency (MDF) in the θ range 36 . In order to confirm our previous findings, we carried out recording of the spontaneous electroencephalogram (EEG) of adult WT and ArhGAP15 −/− animals at two ages: P40 and P180. At P40, the power spectrum of mutant animals (N = 9), compared to the WT animals (N = 10) showed a significant increase of the relative α-(8-12 Hz), of the β-(12-30 Hz) and of the γ-activity (>30 Hz) (Fig. 7A,B). The overall θ-activity (4-8 Hz) appeared unchanged, however when we subdivided the θ-waves in a high (6-8 Hz) and a low (4-6 Hz) range, we detected a significant increase in the θ low activity in ArhGAP15 −/− mice (Fig. 7C). No change of the relative power of the δ-activity was observed. In ArhGAP15 −/− mice we also observed a significant reduction in the variability of the spectral components relative to the β1 activity (Fig. 7A). The MDF changed between groups, with the ArhGAP15 −/− mice showing a slower MDF than the WT controls in the α range (10.3 vs 11.8 Hz) and in the θ range (5.3 vs 6.1 Hz). At P180 the ArhGAP15 −/− mice (N = 3 for both genotypes) showed an increase in the relative power of the α, β, and γ frequency oscillations, similar to what we observed in 6-weeks old animals, while no change of the MDF was detected (Fig. 7D,E). To exclude anomalies of the pyramidal motor pathway or other evident neuromuscular defects, in the 6-weeks old WT and ArhGAP15 −/− mice we examined the peripheral and cortical-to-spinal cord conduction. We did not observe significant changes in the pyramidal conduction relative to the anterior or the posterior limbs (Fig. 7F). All other parameters were found to be in the normal range. In summary, spectral EEG analysis of ArhGAP15 −/− mice shows an increase in the relative power of medium and high frequencies (α, β and γ ranges), a significant increase in the θ low activity (4-6 Hz) and a mild shift of MDF towards lower frequency (α and θ ranges), similar to the mild slowing of background activity seen in most Mowat-Wilson syndrome patients 21,22 . These latter results are in accordance with the reduced complexity of the cortical network previously described.

Discussion
Hereditary forms of ID in humans are caused by mutations in genes coding for regulators of the Rac1 small-GTPase, or components of the Rac-PAK-LIMK main pathway [37][38][39] , and recently mutations in Rac1 have been reported in few ID individuals 1 . All these mutations are of loss-of-function type, thus expected to cause hypo-activity of the Rac-PAK-LIMK transduction pathway and aberrant actin dynamics 40 . How altered actin dynamics result in cognitive deficits in human is still largely unclear, although by examining the roles of different genes in many strains of (loss-of-function) mutant mice, it is generally assumed that cognitive deficits are linked to altered synaptic networking and plasticity, as well as excitation/inhibition imbalance 7,40,41 .
One aspect that has remained poorly studied is whether any misregulation of Rac activity, e.g. hyper-or hypo-, would alter actin dynamics in ways that present common features and thus result in related cognitive deficits. There are substantial similarities with the phenotypes of mice with a conditional loss of Rac1 or of Rac1+ Rac3; specifically the single and combined Rac1/Rac3 mutant animals show defective axonal elongation, altered synaptic density and plasticity, associated with memory and learning deficits 10,13,40 , similar to the ArhGAP15 ones. This observation indicates that the hypo-activity and the hyper-activity of Rac1 can induce similar neuronal defects, and suggests that a fine regulation of the Rac1 activity is necessary to attain the full complexity and functioning of cortical neuronal network.
Considering that, in human, mutations in Oligophrenin1 (a Rho GAP) cause X-linked ID 42,43 , and that the loss of ArhGAP15 (a Rac GAP) is linked to rare variants of the Mowat-Wilson disease with severe neurologic and cognitive deficits 19 , here we examined a mouse model in which Rac1 is hyperactive as the consequence of loss of the Rac-specific GAP protein ArhGAP15. We observe delayed axonal lengthening, reduced neuronal complexity and altered synaptic density on cortical pyramidal neurons, associated with their mild hyperexcitability. We also observe an increase in the relative power in the α, β, and γ ranges, accompanied by cognitive/learning deficits in the adult animals 18 . Frequency oscillations could reflect disturbed neuronal synchronization, e.g., caused by a dysfunctional GABA/glutamate system 44 . Altogether, the synchronous occurrence of θ, α, β and γ activity indicates the existence of distributed oscillatory systems which are linked with sensory and cognitive functions 45 . Of interest, a diminished cortical γ activity has been linked to impaired inhibitory input from fast-spiking parvalbumin-expressing neurons 46 , and we observe indeed a reduced density of GABA synapses. On the contrary, an increased γ activity has been linked to increased excitability of pyramidal neurons which enhances the inhibitory function of parvalbumin neurons via local circuits 34 . In the absence of ArhGAP15 this second mechanism seems prevalent, although the number, distribution and activity of inhibitory neurons in the ArhGAP15 −/− cortex will have to be examined in details.
Recent studies have demonstrated that GAP proteins, by modulating the Rac1 activity, regulate nearly all aspects of brain development, being essential for the regulation of neurogenesis, neuronal morphology, axon growth, synapse formation and plasticity 47,48 . Indeed mutations in GAP molecules (specific for Rho-class GTPases) cause axonal, dendritic and synaptic defects. Rho-class GTPases, control actin cytoskeleton reorganization in complex ways 49 . The best studied pathway involves the PAK1/3-dependent phosphorylation of LIMK and consequent phosphorylation (and inactivation) of ADF/cofilin 29,30 . Protein kinases and phosphatases responsible for cofilin phosphorylation and dephosphorylation play crucial roles in regulating actin cytoskeleton reorganization 50 , and indeed the levels of cofilin phosphorylation drastically change in response to various extracellular stimuli that affect actin dynamics, cell migration and morphogenesis. Here we show that Rac hyperactivity results in a higher level of ADF/cofilin phoshorylation via the increased activity of the PAK-LIMK pathway, and reduced activity of the SSH1L phosphatase. Increased ADF/cofilin phosphorylation, usually associated with reduced actin dynamics 50 , can explain the inefficient neuritogenesis and branching that we observed in the absence of ArhGAP15.  We show that in the absence of ArhGAP15 retrograde actin flow at the growth cone is increased. This should facilitate neurite formation 51 , however this occurs upon coupling of actin retrograde flow with cell adhesion molecules (CAMs) via an efficient linking bridge, known as "clutch" 31,52 . Among the molecules participating in the "clutch", shootin1 is a brain-specific protein that interacts both with the F-actin network and with L1-CAM in axonal growth cones 31,53 . Shootin1 phosphorylation enhances its interaction with the actin network, promoting filopodia and axon extension 32,54,55 . On the contrary, silencing or hypophosphorylation of shootin1 uncouples the actin flow from membrane adhesion and inhibits neurite elongation 32,52,56 . In the absence of ArhGAP15, we observe reduced shootin1 phosphorylation indicating that, in the presence of hyperactive Rac1, shootin1 fails to efficiently couple the F-actin flow with cell adhesion, and this "clutch disengagements" results in shorter and simpler neurites.
Current knowledge on the role of Rac1 for correct neuronal positioning, wiring and connectivity derives largely from the use of either constitutively active (CA) or dominant-negative (DN) Rac1 variants, or the generation of transgenic animals 9,11,15,57 . Such an approach, however, does not provide the ideal experimental setting to study the function of the GTPase. In fact, CA forms require continuous GTP-binding/GTP-hydrolysis cycling for effective signaling flow of the GTPase, while DN forms drastically interfere with endogenous functions. Likewise, the elimination of GTPases, such as their genetic knockouts, alters their amount but not their activity. Not surprisingly in vitro and in vivo results have led to contradictory conclusions. For example, hyperactive cdc42 in vitro promotes dendritic arbor growth 58 , while the deletion of the cdc42-GAP NOMA-GAP in vivo results in hyperactive cdc42 and reduced (not increased) complexity of dendritic arborization 59 . Thus, the approaches of directly interfering or eliminating Rac1 appear inadequate to fully comprehend the physiological role of active vs. inactive GTPases and the neuronal bases of human ID in the presence of mutant GTPase regulators. Instead, manipulating Rac1 regulators offers a better opportunity to alter Rac1 activity in ways that more closely recapitulates the condition of hereditary forms of ID.
Defining the impact of modest misregulation of GTPases activity on the construction, function and plasticity of cortical networks is a main scientific and medical issue, that will require addressing the role of each neuronal-specific GTPase regulator in improved models, such as human iPSC-derived neurons and brain organoids. Relevant information may open ways to pharmacologically modulate Rac activity for therapeutic purposes 40 .

Materials and Methods
Mouse strains. All animal procedures were approved by the Animal Ethics Committee of the University of Torino and by the Ministry of Health. Animals were maintained according to institutional animal welfare guidelines and legislation, under veterinarian surveillance. The ArhGAP15 −/− mouse strain has been previously described 23 . The targeted mutation disrupts the first coding exon by inserting a lacZ reporter cassette and replacing it with the neomycin-resistance cassette. ArhGAP15 +/− mice were backcrossed in C57BL/6 J for five generations before generating homozygous. Heterozygous and homozygous mutant mice are born at normal mendelian frequency, appear overall normal, are viable and fertile, could feed and mate at regular rates, and did not show evident neurological or motor impairments 23 . Animals were maintained in a mixed C57/BL6:DBA genetic background.
Brain preparation for histological analysis. For embryo collection, the day of the vaginal plug was considered as embryonic age 0.5. Embryos were collected by caesarean section from anesthetized pregnant dams, transferred in PBS, and either dissociated for neuronal cultures, or fixed in cold 4% PFA for 8-12 hrs and processed for cryopreservation and histology. Extra-embryonic tissues were used for genotyping by PCR.
For the collection of postnatal and adult brains, WT and ArhGAP15 −/− mice were anesthetized with Avertin (for each animal, 30 μl pure Avertin in 400 μl PBS), transcardially perfused with 5 mM sodium phosphate-buffered 0.9% saline (PBS, pH 7.4) and then with 2% formaldehyde, pH 7.0 (adjusted with NaOH). After the mice were left thus for 2 hrs, their brains were removed, immersed in PBS, and kept in a solution of 30% sucrose in PBS for cryoprotection. The blocks were cut into 25 μM thick coronal sections on a freezing microtome at −20 °C; they were collected in multi-well plates and immersed in PBS.
WT and ArhGAP15 −/− females were mated respectively with WT and ArhGAP15 −/− males; pregnant dams were anesthetized as described above and embryos were collected by caesarean section at the embryonic day 15.5 (E15.5). Embryonic heads were post-fixed overnight; then heads were washed in PBS and kept in a solution of 30% sucrose in PBS for cryoprotection, cut into 20 μm-thick coronal sections and collected on super-adhesive glass slides.
For the histoenzymatic detection of βgal expression, brains specimens were fixed briefly with cold 4% PFA, washed with PBS several times, then included in 4% agarose, vibratome-sectioned (150 μm-thick). Sections were collected on slides and stained with the chromogenic substrate Xgal, according to standard methods. Primary antibodies were incubated in a solution composed by 0.8% BSA in PBS, 2% Goat Serum and PBS 0.1% TritonX-100, overnight at 4 °C, followed by three rinses with PBS for 10 min.

Antibodies and immunostaining.
Sections were incubated for 2 hrs with secondary antibodies in PBS, followed by three rinses with PBS for 10 min. Finally sections were mounted onto coated super-adhesive glass slides and covered with MOWIOL for microscopic observation. For photodocumentation, we used a Leica SP5 confocal microscope with Z-sections of SCIENTIFIC REpoRts | (2018) 8:7254 | DOI:10.1038/s41598-018-25354-3 0.5 μm. Raw images were digitally processed to normalize the background and optimize the contrast, rotated and sized with Photoshop (Adobe).
In situ hybridization. Perfused brains were post-fixed overnight, rinsed in RNAse-free PBS, cryoprotected in 30% sucrose for 24 hrs, embedded/frozen in OCT and sectioned at 20 μM thickness. Hybridization was performed with DIG-labeled riboprobes corresponding to the antisense sequence of the murine ArhGAP15 cDNA or of bacterial lacZ. Sections were permeabilized with 3 μg/ml proteinase K, washed in PBS, and acetylated with 1.3% triethanolamine and 0.25% acetic anhydrate at R.T. Sections were prehybridized in 50% formamide at 60 °C, hybridized with the DIG-labeled probes for 16 hrs, washed, incubated with an anti-DIG-AP antibody (Roche), and developed with NBT-BCIP (Sigma).
Rac1/Rac3 activity was determined by pull-down assays using a glutathione-S-transferase-PAK-CD (PAK-CRIB domain) fusion protein, containing the Rac binding region from human PAK1. Lysates of the embryonic cortices were centrifuged at 4 °C for 10 min at 13,000 RPM, and the supernatant was incubated with glutathione S-transferase PAK glutathione-coupled Sepharose 4B beads (GE Healthcare) for 30 min. at 4 °C. Proteins bound to the beads were washed 3 times in lysis buffer, and then quantified by polyacrylamide gel electrophoresis and Western blot analysis using an anti-Rac1 antibody (Upstate Biotechnology; Lake Placid, USA). An anti-ArhGAP15 monoclonal antibody was also used, this was raised in mouse against a GST-fusion protein of an ArhGAP15 peptide spanning from aminoacids 220-320. Images were quantified by densitometric analysis using Quantity One software (BioRad, CA, USA). The G-LISA colorimetric assay (Cytoskeleton Inc, CO USA) was used according to the manufacturer's instructions.
Primary cultures of cortical neurons. WT, ArhGAP15 +/− or ArhGAP15 −/− embryos at the age E15.5 were used to establish primary cultures of cortical neurons. Embryonic heads were dissected in sterile conditions in cold Leibovitz's L-15 Medium (Gibco, Life Technologies), the cortices were dissected free of the rest of the brain, deprived of the meninges and dissociated. For the dissociation, Neurobasal B27 (Neurobasalmedium, Glutamine 1/100, B27 1/50, Gentamicine 1/1000) was used, first with mechanical shearing and then by adding trypsin for 15 minutes at 37 °C, followed by centrifugation and resuspension in culture medium. Two cellular pools, WT and ArhGAP15 −/− , were thus obtained. Each pool of cell suspension was quantified through Countess (Life Technologies) and 7 × 10 6 cells from each pool was nucleofected (i.e. electroporated) with PGK-eGFP vector (2 μg) or with the LifeAct-RFP vector (2 μg) 28 using Nucleofector 2b (Nucleofector System, Lonza, Basel, Switzerland) with the electroporation code O-005 61 .
Nucleofected neurons were plated on round glasses placed in 3.5 cm diameter Petri dishes, previously coated with poly-L-lysine (1 mg/ml; Sigma; 1/10 dilution) at the density of 10 6 /cm 2 , and allowed to adhere. After cells attach to the glasses, medium was replaced with fresh plating medium to remove unattached cells and neurons were incubated for 3 days at 37 °C in 5% CO 2 saturation atmosphere. Cells were observed during the three days in order to verify their culture condition and the absence of contaminations; then culture medium was aspirated and cells were washed with sterile PBS; cells were fixed for 20 min. with PFA 4% and washed again in PBS. Round glasses with adherent cortical primary neurons were picked up with tweezers, laid on microscopy glass, then mounted with MOWIOL and photographed.
Fluorescence images were acquired using an inverted microscope (Axio Observed Z1, Zeiss), ApoTome system. Images were digitally captured using a cooled 16-bit camera (Axio MRM, Zeiss) with Axio Vision Release 4.7.1 software.
Morphological analyses. Cultured WT and ArhGAP15 −/− neurons were classified for their morphological features as unipolar, bipolar and multipolar, based on the presence of 1, 2 or >2 main neurites stemming from the cell body, as shown in Supplementary Figure S2. Analyses were performed with the ImageJ informatics suite, and the ImageJ plugin Cell Counter. Cultured WT and ArhGAP15 −/− neurons were then analyzed for the main neurite length, the number of secondary neuritis, and the number and distribution of branches (indicating the arborization rate): together these criteria define the complexity of neuronal structure. A minimum of 150 WT and ArhGAP15 −/− neurons were counted. To perform this analysis, we used the basic functions of ImageJ software 62 , measuring the length of the main neurite of each neuron and the number of secondary neurites, departing from the main neurite. Arborization of each neuron was quantified by Sholl analysis 63 , a computer-assisted method uses concentric circles around the neuronal soma, and determines how many times each neuron intersects each of these circles (plugin ImageJ).
The same methods were applied to examine fluorescent images of individual neurons stained with the DiI retrograde tracing method (see below). For these analyses, at least 50 neurons belonging to the same cortical area and layer were traced for each genotype with the Neurolucida software package. Neurons were analyzed using the Neurolucida Explorer software package (MBF Bioscience) quantifying the length of axon, dendrites and apical dendrite, and making the Sholl analysis.
Anterograde tracing of cortical callosal axons. We used a procedure described 64 . WT and ArhGAP15 −/− animals at the age P2 were transcardially perfused with 2% PFA, removed from the skull and post-fixed for at least 7 days. Crystals of 1, 1′-Dioctadecyl-3, 3, 3′, 3′-Tetramethylindocarbocyanine Perchlorate (DiI, size 25 μm) (from Molecular Probes) were inserted with a fine needle in the right somatosensory cortex (AP = +0.8; ML = +1.5; DV = −0.5), and the dye was allowed to diffuse for 7 days at 37 °C. Mice at the age P8 were subjected to the same procedure, with the exception of the diffusion time which was extended to 14 days. Thick (100 μm) vibratome sections were then examined with a fluorescent microscopy. Images were used to computer assisted analyses to estimate the maximum and the average length of the callosal axons. Seven and 12 sections (from at least 3 independent brains) were examined for, respectively, the WT and the ArhGAP15 mutant genotypes. The fluorescence intensity (F.I.) and the thickness (T) of the callosal bundle of axons was measured at various distances (from 0 to 1612 μm) from the callosal commissure, set as distance = 0. The multiplication F.I.*T was calculated at each distance (F.I.*T-d), and the results were expressed as F.I.*T-d relative to the F.I.*T at distance = 0. The length of the five longest axons was also determined, on a minimum of 50 axons from at least 3 independent brain samples for each genotype. At the age P45, cortical neurons were labeled by pressure injecting 0.5 μL of Biotinylated Dextran Amine (BDA, Thermo Fisher) MW 10 000 (5% in 0.01 M phosphate buffer) 65 into the right cortex of anesthetized WT and ArhGAP15 −/− animals, the mice were let recover and maintained in standard conditions for 7 days. Animals were then transcardially perfused, the brains were collected and cryosectioned (40 μM thick). Floating sections were stained with avidin-biotinylated HRP procedure and standard DAB reaction, and mounted on glass slides.
Time-Lapse video recordings. Dissociated cortical neurons from E15.5 WT and ArhGAP15 −/− embryos were prepared and nucleofected with the Life-act vector. After 2 DIV individual growth cones were examined by fluorescent time-lapse video. Time lapses were recorded for 5 minutes with an interval of 10 seconds using a 40X PlanApo N.A. 1.4 oil immersion objective on the cells kept in the microscope incubator at 37 °C and 5% CO 2 . The acquired movies (N = 60 and 65 good quality ones, for WT and ArhGAP15 −/− , respectively) were used to determine the total anterograde and retrograde actin movements, using ImageJ dedicated plugins. Briefly, for each growth cone a kimograph was first extracted, then the movements parameters were computed and exported for statistical analyses.
Current-clamp analysis were performed on pyramidal neurons of the somatosensory cortex, identified by their morphological features, under video microscopy with an internal solution containing (in mM) 126 K-gluconate, 4 NaCl, 1 EGTA, 1 MgSO 4 , 0.5 CaCl 2 , 3 ATP (magnesium salt), 0.1 GTP (sodium salt), 10 glucose, 10 HEPES-KOH (pH 7.28; osmolarity adjusted to 280 mOsm). A series of current steps (from −60 pA to 240 pA) were injected to induce action potentials (20-pA injection current per step, duration of 1 s) in order to study passive and active properties of the neuron membrane. Data analysis was performed offline with Clampfit 10.1 software. EEG recordings. Three-months old WT and ArhGAP15 −/− females (in C57BL6) were maintained in the same environmental conditions, then anaesthetized with Avertin and used for EEG recording. Two Ag-AgCl ball electrode were placed over the left neocortex, the common reference electrode was fixed on the right pinna. The recording of the ECG activity was performed using a 16-channel data acquisition system (PowerLab 16/30, AD Instruments, Australia). The biological signals from the electrodes were amplified and filtered (0.1-50 Hz bandpass) using BioAmp amplifiers (AD Instruments). The EcoG signal was digitalized at a sampling rate of 1024, then displayed and stored. Only the spikes with amplitude greater than three-fold the baseline were included. The EEG power spectrum was analysed with LabChart 7.2 software (AD Instruments). FFT analysis was applied for at least 5 min time windows of continuous noise-free segments. FFTs were collected as power spectra with a sample size of 2048 (without overlap). EEG power was computed in the selected frequency ranges: δ range (0.5-4 Hz), θ range (4-8 Hz), α range (8-12 Hz), β range (12-30 Hz) and γ range . To normalize the data, EEG power (μV 2 ) for each frequency bin within each epoch was calculated as a percentage of the total EEG power (1-56 Hz) of all epochs.